Conversational AI Impacts Learning
(A2026-134027)

AUTHORS

Deniz Lefkeli, LUISS Guido Carli University; Ada Aka, Stanford University; Konstantin Pikal, Kedge Business School

ABSTRACT

Conversational AI has rapidly gained popularity as a tool for acquiring information; however, little is known about its effect on learning. Across 9 studies, we show that conversational language impairs learning despite being more preferred for informational queries. Studies 1A-1C demonstrated a relative preference for conversational language for informational searches, but not for commercial, navigational, or transactional searches. However, Studies 2A-2C revealed that conversational language impaired the adoption of a learning mindset and recall performance. Studies 3A-B showed that conversational language decreased perceived informational value, impairing learning. Study 4 tested three debiasing interventions (i.e., bias-informing, source-of-bias, and consider-the-opposite), but none successfully mitigated the learning impairment. Study 5 replicated the recall deficit in a more naturalistic setting where people used either conversational AI or traditional search engines. These findings reveal that conversational language can make informational searches appealing; however, it may simultaneously undermine learning outcomes.